webhook: Retry OSError on queue connection failure

Commit 06c2717 introduced a retry loop for connecting to AMQP, as
these two containers start at the same time in the OpenShift deployment.

However, these days the webhook often fails with `ssl.SSLEOFError`
(with parents `SSLError` and `OSError`), which doesn't fit `AMQPError`.
So intercept this as well.

With that we can revert the static sleep from commit 169b53b
again.

Fixes #635

tasks/cockpit-tasks-webhook.yaml

@@ -39,7 +39,7 @@ spec:
           ports:
             - containerPort: 8080
               protocol: TCP
-          command: [ "sh", "-ec", "sleep 10; exec webhook" ]
+          command: [ "webhook" ]
           volumeMounts:
           - name: webhook-secrets
             mountPath: /run/secrets/webhook

tasks/container/webhook

@@ -88,7 +88,7 @@ def scan_tasks():
         try:
             with distributed_queue(AMQP_SERVER):
                 break
-        except pika.exceptions.AMQPError as e:
+        except (OSError, pika.exceptions.AMQPError) as e:
             logging.info("Failed to connect to AMQP, attempt #%i: %s", retry, e)
             time.sleep(5)